Whamcloud - gitweb
LU-1252 recovery: reduce reconnect time for IR
authorJinshan Xiong <jinshan.xiong@whamcloud.com>
Sat, 19 May 2012 16:45:23 +0000 (09:45 -0700)
committerOleg Drokin <green@whamcloud.com>
Tue, 22 May 2012 18:32:35 +0000 (14:32 -0400)
commitde936be08482bd55ceed8e741119d9a9ce4746b2
tree5cf3779aee421d33dbe9ae1d6fc6e49ecf632f25
parent632cef2dcd807b1327ee26aacd9698953fb62c5b
LU-1252 recovery: reduce reconnect time for IR

Two problems fixed in this patch:
1. network latency can be increased to incredible large after all
   connections have been tried. Limited it to not be over
   CONNECTION_SWITCH_MAX;
2. reconnect the failing target ASAP when the previous connecting
   RPC failed, if the client has already been notified by IR.
With the above two fix, it can reduce the recovery time if there was a
connecting RPC in flight when the client is notified by IR.

Signed-off-by: Jinshan Xiong <jinshan.xiong@whamcloud.com>
Change-Id: If43e93037e418b7a775228ca7abbe8b337e44e9a
Reviewed-on: http://review.whamcloud.com/2371
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
lustre/include/lustre_import.h
lustre/ptlrpc/import.c